Beyond virtualization: Envision Real Cloud computing

I find it very confusing to see a modest IT infrastructure with little or no virtualization at all, and I'm even more surprised when I learn that they have no intention of accepting virtualization in the near future. Whether it's out of the "get out of my way" attitude or simply to reduce the budget, it is crazy to stick to the traditional physical infrastructure today.

First, what do these companies buy as servers? If they replace old single and dual-core servers with new four-or-quad servers and simply move services, they already have enough hardware. Each server workload runs on hardware that can easily handle 6 virtual servers, even if it includes a free system management program. Are these companies accepting virtualization when server virtualization becomes a past?

Looking to the next few years, we can expect to see a fundamental change in how the virtual server is managed. As if the enterprise virtualization from beginning to end is not a real revolution, I think there will be a lot of storms will be swept in the future. The next leap will occur when an operating system that is explicitly designed and tuned to run as a VM appears, because of the lack of meaningful driver support and the nature of other physical server worlds, this operating system will not even run on physical hardware. What's the point of bringing thousands of drivers and hardware from a particular feature into the VM? They have no use at all!

Ultimately, we'll see a widely used version of the operating system: it replaces most of today's supported operating systems, solves years of redundancy and bloat to support the kernel dedicated to managing programs, and has a unique idea of how memory, CPU, and I/O resources work and manage. Now that we're back to the era of introducing CPU and RAM into the VM through hot-swappable extensions, this hot-plug expansion was originally conceived as an extension for physical RAM and CPUs.

But there is still a long way to go before the idea can be realized. When an operating system kernel can instantly request, adapting to and using additional computational resources without having to understand the information behind those resources, we are closer to the concept of true cloud computing: no underlying hardware, no fixed CPU concept, no static or fixed RAM concepts in each operation. These operating systems are tightly integrated through the scheduling hierarchy of management programs, with little concern for the affinity of a single processor core, or the optimization of NUMA or running RAM.

It will take many years to fully realize it, but I am sure it is coming. By then, the hypervisor will become strong and smart enough to manage functions for any compatible host OS, essentially turning each VM into an application pool. As the application load increases, the kernel is more concerned with the resource requirements of the growing management program, which then makes lower-level decisions that determine what physical resources the VM can consume and, if necessary, transfer other instances to different physical servers to free up space.

This, in turn, removes the notion of specifying the analog CPU and RAM limits rather than setting the minimum, maximum, and burst limits to the VM-similar to parallel virtualization, but completely unrelated to the OS, and does not rely on the kernel host OS as a middle-tier and fixed kernel. Each VM may know that it is a running VM, but managing transactions through a hypervisor also runs its own kernel and solves hardware problems directly.

As the ability to upgrade in operation occurs, that is, by upgrading the passive management program, the load near the immediate transfer, and then upgrading other aspects, we say that a fixed-use server with embedded management programs may be redundant, as is the case for small physical servers, These small physical servers are basically composed of CPUs and RAM slots that contain a fully software-managed Backplane-level network interface.

Are these advanced virtual servers completely virtual? Or are they completely servers? If they develop into software services that resemble databases and applications that understand their environment, there is no need to run multiple instances at the same time.

You might have two instead of dozens of Web service VMS, but these two VM frequencies can grow from 500MHz to 32GBHz, because the load jumps from 6GB RAM to 512GB at the same time without assigning any fixed values. As you can imagine, after using the appropriate high-speed backplane, as the management program partitions priority tasks, keeping RAM from local to a single process, a server instance can span two or more physical servers. Of course, we're talking about a whole multithreaded application, but isn't that the whole idea behind virtualization and multi-core CPUs?

Perhaps, one day in the future, we will see one or even several blade chassis, we see not dozens of of the above running the management program of the physical server, instead, we think of them as a lot of services will be consumed by any number of resources. If a hardware failure occurs, it is easy to replace the faulty part by Hot swappable, the only loss may be that hundreds of processes disappear from the tables that contain tens of thousands of processes, and these disappearing processes will be restarted elsewhere in an instant, without significant loss of service.

Perhaps, if we can see such a reality, those stubborn people who are still maintaining a dilapidated physical datacenter may end up endorsing the idea of running multiple servers in each physical enclosure. By then, they will have no choice, and there will be no alternative.